AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Spread Spurious Attribute: Improving Worst-group Accuracy with Spurious Attribute Estimation

Junhyun Nam, Jaehyung Kim, Jaeho Lee, Jinwoo Shin

ICLR 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Group-robust classification · 11 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use SSA with pseudo-attribute estimation and Group DROworst-group robust training · Waterbirdsworst-group accuracy: 86.7 -> 89.0Improved
Use SSA with pseudo-attribute estimation and Group DROworst-group robust training · CelebAworst-group accuracy: 81.1 -> 89.8Improved
Use SSA with pseudo-attribute estimation and Group DROworst-group robust training · MultiNLIworst-group accuracy: 72.6 -> 76.6Improved
Use SSA with pseudo-attribute estimation and Group DROworst-group robust training · CivilComments-WILDSworst-group accuracy: 69.3 -> 69.9Improved
Add fixed-threshold pseudo-labeling (ablation)spurious-attribute predictor · CelebA, Blond/Male group, 10% validation labelsgroup accuracy: 70.4 -> 76.4Improved
Add fixed-threshold pseudo-labeling (ablation)spurious-attribute predictor · CelebA, Blond/Male group, 5% validation labelsgroup accuracy: 68.7 -> 67.8Hurt
Add group-wise adaptive thresholds (ablation)pseudo-labeling · CelebA, Blond/Male group, 10% validation labelsgroup accuracy: 76.4 -> 83.6Improved
Add group-wise adaptive thresholds (ablation)pseudo-labeling · CelebA, Blond/Male group, 5% validation labelsgroup accuracy: 67.8 -> 75.7Improved
Split group-unlabeled samples into predictor-training and prediction subsets (ablation)SSA pseudo-labeling · CelebA, Blond/Male group, 10% validation labelsgroup accuracy: 76.3 -> 83.6Improved
Split group-unlabeled samples into predictor-training and prediction subsets (ablation)SSA pseudo-labeling · CelebA, Blond/Male group, 5% validation labelsgroup accuracy: 71.7 -> 75.7Improved
Split group-unlabeled samples into predictor-training and prediction subsets (ablation)SSA pseudo-labeling · Waterbirds, Landbird/Land group, 10% validation labelsgroup accuracy: 91.3 -> 92.1Improved
Split group-unlabeled samples into predictor-training and prediction subsets (ablation)SSA pseudo-labeling · Waterbirds, Waterbird/Land group, 10% validation labelsgroup accuracy: 92.9 -> 94.6Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy